Interface StrategyFunction

All Known Subinterfaces:

StrategyFunction.Deterministic, StrategyFunction.NonDeterministic

public interface StrategyFunction

The heuristic function for determining the strategy at the root of any B* search. This class contains several defaults for strategy functions, but you can also construct your own using this interface and several methods for constructing new strategy functions.

Nested Class Summary

Nested Classes

Modifier and Type	Interface	Description
static interface	StrategyFunction.Deterministic
static interface	StrategyFunction.HeuristicFunction
static interface	StrategyFunction.NonDeterministic

Field Summary

Fields

Modifier and Type	Field	Description
static final StrategyFunction.Deterministic	B_2D	Berliner's depth-based D-criterium, considering 1 alternative
static final StrategyFunction.Deterministic	B_2R	Berliner's depth- and range-based R-criterium, considering 1 alternative
static final StrategyFunction.Deterministic	B_3D	Berliner's depth-based D-criterium, considering 2 alternatives
static final StrategyFunction.Deterministic	B_3R	Berliner's depth- and range-based R-criterium, considering 2 alternatives
static final StrategyFunction.Deterministic	B_AD	Berliner's depth-based D-criterium, considering all alternatives
static final StrategyFunction.Deterministic	B_AR	Berliner's depth- and range-based R-criterium, considering all alternatives
static final StrategyFunction.Deterministic	DISPROVEREST	will use DISPROVEREST where possible
static final StrategyFunction.Deterministic	P_2M1
static final StrategyFunction.Deterministic	P_2M2
static final StrategyFunction.Deterministic	P_2S1
static final StrategyFunction.Deterministic	P_2S2
static final StrategyFunction.Deterministic	P_2X1
static final StrategyFunction.Deterministic	P_2X2
static final StrategyFunction.Deterministic	P_3M1
static final StrategyFunction.Deterministic	P_3M2
static final StrategyFunction.Deterministic	P_3S1
static final StrategyFunction.Deterministic	P_3S2
static final StrategyFunction.Deterministic	P_3X1
static final StrategyFunction.Deterministic	P_3X2
static final StrategyFunction.Deterministic	P_AM1
static final StrategyFunction.Deterministic	P_AM2
static final StrategyFunction.Deterministic	P_AS1
static final StrategyFunction.Deterministic	P_AS2
static final StrategyFunction.Deterministic	P_AX1
static final StrategyFunction.Deterministic	P_AX2
static final StrategyFunction.Deterministic	PROVEBEST	will ALWAYS use PROVEBEST

Method Summary

Modifier and Type	Method	Description
static StrategyFunction.NonDeterministic	ALTERNATE()	will alternate between PROVEBEST and DISPROVEREST, starting at PROVEBEST; unless DISPROVEREST is impossible, which will suspend the alternation count
static StrategyFunction.Deterministic	BY_SUBTREE_EXPANSIONS(int num_alternatives, boolean by_optimistic, double best_node_weight)
static StrategyFunction.Deterministic	BY_SUBTREE_MAXSIZE(int num_alternatives, boolean by_optimistic, double best_node_weight)
static StrategyFunction.Deterministic	BY_SUBTREE_PARAMS(int num_alternatives, boolean by_optimistic, ToDoubleFunction<MetricKeeper> bestNodeEvaluator, ToDoubleFunction<MetricKeeper> restNodesEvaluator)
static StrategyFunction.Deterministic	BY_SUBTREE_SIZE(int num_alternatives, boolean by_optimistic, double best_node_weight)
static StrategyFunction.Deterministic	CRITERIUM_D_or_R(int num_alternatives, boolean divide_by_range)	If the sum of the squares of the depths from which the optimistic bounds of the alternatives has been backed up is less than the square of the depth from which the value of the best arc has been backed up, then use DISPROVEREST; otherwise use PROVEBEST.
static StrategyFunction.Deterministic	getHeuristicStrategyFunction(int num_alternatives, boolean by_optimistic, StrategyFunction.HeuristicFunction bestNodeFunction, StrategyFunction.HeuristicFunction restNodesFunction)	If the sum of the restNodesFunction values for the alternatives is less than the bestNodeFunction value for the best node, then use DISPROVEREST; otherwise use PROVEBEST.
boolean	isDeterministic()
static StrategyFunction.NonDeterministic	RANDOM()	will select randomly between PROVEBEST and DISPROVEREST when both are applicable
static StrategyFunction.NonDeterministic	RANDOM(long seed)	will select randomly between PROVEBEST and DISPROVEREST when both are applicable
boolean	useProveBest(GameTreeNode<?,?> root, MetricKeeper... metrics)

Field Details

PROVEBEST

static final StrategyFunction.Deterministic PROVEBEST

will ALWAYS use PROVEBEST

DISPROVEREST

static final StrategyFunction.Deterministic DISPROVEREST

will use DISPROVEREST where possible

B_2D

static final StrategyFunction.Deterministic B_2D

Berliner's depth-based D-criterium, considering 1 alternative

B_3D

static final StrategyFunction.Deterministic B_3D

Berliner's depth-based D-criterium, considering 2 alternatives

B_AD

static final StrategyFunction.Deterministic B_AD

Berliner's depth-based D-criterium, considering all alternatives

B_2R

static final StrategyFunction.Deterministic B_2R

Berliner's depth- and range-based R-criterium, considering 1 alternative

B_3R

static final StrategyFunction.Deterministic B_3R

Berliner's depth- and range-based R-criterium, considering 2 alternatives

B_AR

static final StrategyFunction.Deterministic B_AR

Berliner's depth- and range-based R-criterium, considering all alternatives

P_2M1

static final StrategyFunction.Deterministic P_2M1

P_2S1

static final StrategyFunction.Deterministic P_2S1

P_2X1

static final StrategyFunction.Deterministic P_2X1

P_2M2

static final StrategyFunction.Deterministic P_2M2

P_2S2

static final StrategyFunction.Deterministic P_2S2

P_2X2

static final StrategyFunction.Deterministic P_2X2

P_3M1

static final StrategyFunction.Deterministic P_3M1

P_3S1

static final StrategyFunction.Deterministic P_3S1

P_3X1

static final StrategyFunction.Deterministic P_3X1

P_3M2

static final StrategyFunction.Deterministic P_3M2

P_3S2

static final StrategyFunction.Deterministic P_3S2

P_3X2

static final StrategyFunction.Deterministic P_3X2

P_AM1

static final StrategyFunction.Deterministic P_AM1

P_AS1

static final StrategyFunction.Deterministic P_AS1

P_AX1

static final StrategyFunction.Deterministic P_AX1

P_AM2

static final StrategyFunction.Deterministic P_AM2

P_AS2

static final StrategyFunction.Deterministic P_AS2

P_AX2

static final StrategyFunction.Deterministic P_AX2

Method Details

useProveBest

boolean useProveBest(GameTreeNode<?,?> root,
 MetricKeeper... metrics)

isDeterministic

boolean isDeterministic()

ALTERNATE

static StrategyFunction.NonDeterministic ALTERNATE()

will alternate between PROVEBEST and DISPROVEREST, starting at PROVEBEST; unless DISPROVEREST is impossible, which will suspend the alternation count

RANDOM

static StrategyFunction.NonDeterministic RANDOM()

will select randomly between PROVEBEST and DISPROVEREST when both are applicable

RANDOM

static StrategyFunction.NonDeterministic RANDOM(long seed)

will select randomly between PROVEBEST and DISPROVEREST when both are applicable

CRITERIUM_D_or_R

static StrategyFunction.Deterministic CRITERIUM_D_or_R(int num_alternatives,
 boolean divide_by_range)

If the sum of the squares of the depths from which the optimistic bounds of the alternatives has been backed up is less than the square of the depth from which the value of the best arc has been backed up, then use DISPROVEREST; otherwise use PROVEBEST.

Parameters:

num_alternatives - number of alternatives (including best arc) to consider. Or any value <1 to consider ALL.

divide_by_range - toggle the use of the "R" criterium, where the square of the depth value is divided by the node's range before summing and comparing them.

Returns:

a function which follows the described behavior for the selected number of alternatives considered to determine the strategy. This function will return true when PROVEBEST should be used, and false otherwise.

BY_SUBTREE_MAXSIZE

static StrategyFunction.Deterministic BY_SUBTREE_MAXSIZE(int num_alternatives,
 boolean by_optimistic,
 double best_node_weight)

BY_SUBTREE_SIZE

static StrategyFunction.Deterministic BY_SUBTREE_SIZE(int num_alternatives,
 boolean by_optimistic,
 double best_node_weight)

BY_SUBTREE_EXPANSIONS

static StrategyFunction.Deterministic BY_SUBTREE_EXPANSIONS(int num_alternatives,
 boolean by_optimistic,
 double best_node_weight)

BY_SUBTREE_PARAMS

static StrategyFunction.Deterministic BY_SUBTREE_PARAMS(int num_alternatives,
 boolean by_optimistic,
 ToDoubleFunction<MetricKeeper> bestNodeEvaluator,
 ToDoubleFunction<MetricKeeper> restNodesEvaluator)

getHeuristicStrategyFunction

static StrategyFunction.Deterministic getHeuristicStrategyFunction(int num_alternatives,
 boolean by_optimistic,
 StrategyFunction.HeuristicFunction bestNodeFunction,
 StrategyFunction.HeuristicFunction restNodesFunction)

If the sum of the restNodesFunction values for the alternatives is less than the bestNodeFunction value for the best node, then use DISPROVEREST; otherwise use PROVEBEST.

Parameters:

num_alternatives - number of alternatives (including best node) to consider. Or any value <1 to consider ALL.

by_optimistic - if true, consider the nodes with better optimistic values to be 'better'. If false, instead consider the nodes with better pessimistic values to be 'better'. This ordering of nodes affects which node is selected as 'best' node and which nodes are selected as alternative nodes if num_alternatives is not considering ALL alternative nodes.

bestNodeFunction - function to apply to the best node

restNodesFunction - function to apply to the alternative nodes

Returns:

a function which follows the described behavior for the selected number of alternatives considered to determine the strategy. This function will return true when PROVEBEST should be used, and false otherwise.